| Inferring opinions based on learned probabilities -> Monitor Keywords |
|
|
  Inferring opinions based on learned probabilitiesInferring opinions based on learned probabilities description/claimsThe Patent Description & Claims data below is from USPTO Patent Application 20080097758, Inferring opinions based on learned probabilities. Brief Patent Description - Full Patent Description - Patent Application Claims
BACKGROUND [0001]Electronic commerce via the World Wide Web ("web") is becoming an increasingly popular way for people to buy products. The people who use the web to buy products often turn to the web for help in deciding what products to buy and from which web retailer. Many sources of product reviews are available via the web to help a person. These product reviews may be provided by professional product review web sites (e.g., CNet), commercial web sites (e.g., Amazon.com), discussion boards, and personal web sites and web logs ("blogs"). A person can use any of these sources of product reviews to help in their buying decision. [0002]Product reviews may be generated by experts or by customers of retail web sites. A professional product review web site may enlist the services of various experts to review products or services that may include movies, music, books, automobiles, electronic products, software, and so on. These experts review the products and provide their opinion on the product via a product review. Many of these professional product review web sites generate significant revenue from advertisements that are presented along with their reviews. To increase traffic to their web sites, these professional product review web sites typically try to ensure the quality of their reviews. In contrast, some retail web sites allow any customer to submit a product review, but may exert no control over the quality and accuracy of the product reviews. [0003]In addition to helping potential buyers make buying decisions, these product reviews may provide valuable feedback to manufacturers who seek to improve the quality of their products. The product reviews provide a wealth of information relating to what experts and customers like and dislike about a manufacturer's products. Because of the large volume of product reviews being created, it can be very difficult and time-consuming for a manufacturer to identify all the product reviews for the manufacturer's products and then to categorize the product reviews as expressing a positive or negative opinion about the product. [0004]Although some attempts have been made to classify product reviews as being positive or negative, these attempts typically try to classify product reviews by applying text classification techniques. These attempts generate training data by classifying product reviews as being positive or negative. The attempts then extract features and train a classifier to classify product reviews based on the features of the reviews. Text classification techniques, however, may not be particularly effective at classifying product reviews. Text classification techniques rely, in large part, on term frequency to identify the topics of a document. Since the opinion of a product review may be expressed clearly only in a single sentence of a long product review, the use of term frequency may not help identify the opinion. Also, some product reviews may have their opinions expressed indirectly or may even attempt to mask their opinion. In such cases, text classification techniques will likely not be able to correctly classify the opinions expressed by the product reviews. SUMMARY [0005]A method and system for determining an opinion expressed via target words is provided. An opinion system determines the opinion expressed by a sentence based on opinion data representing statistical properties of product reviews that express an opinion. The opinion system learns the statistical properties from training data. The training data includes sentences and the opinion (e.g., positive or negative) expressed by each sentence. The opinion system generates the statistics by first identifying sequences of parts of speech or sequential patterns of the training data that are commonly used to express an opinion. The opinion system then calculates from the training data the probabilities that each identified sequence of the parts of speech may contain a certain opinion word. An opinion word is a word that is often used to express an opinion. To complete the generating of the statistics, the opinion system calculates the probabilities from the training data that various opinion words are used to express an opinion. [0006]After the opinion system learns the opinion data, the opinion system can then infer opinions of product reviews using the opinion data. The opinion system infers the opinion of a sentence of a product review based on a probability that the sentence contains certain sequences of parts of speech that are commonly used to express an opinion as indicated by the training data and the probabilities of the training data. When provided with the sentence, the opinion system identifies possible sequences of parts of speech of the sentence. The opinion system may apply natural language processing techniques to identify the sequences along with a probability that each sequence is the correct sequence for the sentence. For each sequence, the opinion system then retrieves from the opinion data the probability that the sequence contains an opinion word. The opinion system then retrieves from the opinion data the probability that the opinion words of the sentence are used to express an opinion. The opinion system then combines the probabilities to generate an overall probability that the sentence with that sequence expresses an opinion. [0007]This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter. BRIEF DESCRIPTION OF THE DRAWINGS [0008]FIG. 1 is a block diagram that illustrates components of the opinion system in one embodiment. [0009]FIG. 2 is a block diagram that illustrates data of the opinion system in one embodiment. [0010]FIG. 3 is a flow diagram that illustrates the processing of the generate opinion data for opinions component of the opinion system in one embodiment. [0011]FIG. 4 is a flow diagram that illustrates the processing of the calculate opinion data component of the opinion system in one embodiment. [0012]FIG. 5 is a flow diagram that illustrates the processing of the identify opinion sequential patterns component of the opinion system in one embodiment. [0013]FIG. 6 is a flow diagram that illustrates the processing of the calculate probability of an opinion word given a sequential pattern component of the opinion system in one embodiment. [0014]FIG. 7 is a flow diagram that illustrates the processing of the calculate probability of an opinion given an opinion word component of the opinion system in one embodiment. [0015]FIG. 8 is a flow diagram that illustrates the processing of the infer opinion component of the opinion system in one embodiment. [0016]FIG. 9 is a flow diagram that illustrates the processing of the calculate probability of opinion given a sentence component of the opinion system in one embodiment. DETAILED DESCRIPTION [0017]A method and system for determining an opinion expressed via target words is provided. In one embodiment, an opinion system determines the opinion expressed by a sentence (i.e., target words) based on opinion data representing statistical properties of product reviews that express an opinion. The opinion system learns the statistical properties from training data. The training data includes sentences and the opinion (e.g., positive or negative) expressed by each sentence. The opinion system generates the statistics by first identifying sequences of parts of speech or sequential patterns of the training data that are commonly used to express an opinion. For example, sentences that express a positive opinion may commonly include a sequence of a noun and a verb followed by an adjective. The opinion system then calculates from the training data the probabilities that each identified sequence of the parts of speech may contain a certain opinion word. An opinion word is a word that is often used to express an opinion. For example, "love," "hate," and "best" are opinion words. These opinion words may likely be used to express an opinion when used with certain sequences of parts of speech. To complete the generating of the statistics, the opinion system calculates the probabilities from the training data that various opinion words are used to express an opinion. For example, the opinion word "best" may or may not be used in product reviews to express an opinion. The sentence "We hired the best experts to evaluate the car" does not express an opinion about the car, whereas the sentence "It is one of the best cars of the year" does express an opinion about the car. In contrast, the opinion word "love" is typically used to express a positive opinion. The opinion system may generate one set of opinion data from sentences that express positive opinions and another set of opinion data from sentences that express negative opinions. After the opinion system learns the opinion data, the opinion system can then infer opinions of product reviews using the opinion data. [0018]In one embodiment, the opinion system infers the opinion of a sentence of a product review based on a probability that the sentence contains certain sequences of parts of speech that are commonly used to express an opinion as indicated by the training data and the probabilities of the training data. When provided with the sentence, the opinion system identifies possible sequences of parts of speech of the sentence. The opinion system may apply natural language processing techniques to identify the sequences along with a probability that each sequence is the correct sequence for the sentence. For each sequence, the opinion system then retrieves from the opinion data the probability that the sequence contains an opinion word. The opinion system then retrieves from the opinion data the probability that the opinion words of the sentence are used to express an opinion. The opinion system then combines the probabilities to generate an overall probability that the sentence with that sequence expresses an opinion. The opinion system may generate, for each sequence of parts of speech, an overall probability that the sentence expresses a positive opinion using the positive opinion data and an overall probability that the sentence expresses a negative opinion using the negative opinion data. The opinion system then selects the opinion associated with the higher probability as the opinion of the sentence. The opinion system may derive the overall opinion of the product review from the opinions of its sentences. [0019]In one embodiment, the opinion system represents the probability that a sentence expresses a positive or negative opinion by the following: P(o|s)=P(o|ow)*P(ow|sp)*P(sp|s) Continue reading about Inferring opinions based on learned probabilities... Full patent description for Inferring opinions based on learned probabilities Brief Patent Description - Full Patent Description - Patent Application Claims Click on the above for other options relating to this Inferring opinions based on learned probabilities patent application. ###  How KEYWORD MONITOR works... a FREE service from FreshPatents1. Sign up (takes 30 seconds). 2. Fill in the keywords to be monitored. 3. Each week you receive an email with patent applications related to your keywords. Start now! - Receive info on patent apps like Inferring opinions based on learned probabilities or other areas of interest. ### Previous Patent Application: Audio coding Next Patent Application: Method and apparatus for remote control in portable terminal Industry Class: Data processing: speech signal processing, linguistics, language translation, and audio compression/decompression ### FreshPatents.com Support Thank you for viewing the Inferring opinions based on learned probabilities patent info. IP-related news and info Results in 2.18483 seconds Other interesting Feshpatents.com categories: Computers: Graphics , I/O , Processors , Dyn. Storage , Static Storage , Printers |
||
